EXAMINERS’ REPORT

CS1B – Actuarial Statistics

Core Principles
Paper B

September 2022
 CS1B – Actuarial Statistics – Core Principles – September 2022 - Examiners’ report

Introduction

The Examiners’ Report is written by the Chief Examiner with the aim of helping candidates,
both those who are sitting the examination for the first time and using past papers as a
revision aid and also those who have previously failed the subject.

The Examiners are charged by Council with examining the published syllabus. The
Examiners have access to the Core Reading, which is designed to interpret the syllabus, and
will generally base questions around it but are not required to examine the content of Core
Reading specifically or exclusively.

For numerical questions the Examiners’ preferred approach to the solution is reproduced in
this report; other valid approaches are given appropriate credit. For essay-style questions,
particularly the open-ended questions in the later subjects, the report may contain more points
than the Examiners will expect from a solution that scores full marks.

The report is written based on the legislative and regulatory context pertaining to the date that
the examination was set. Candidates should take into account the possibility that
circumstances may have changed if using these reports for revision.

Sarah Hutchinson
Chair of the Board of Examiners
December 2022

CS1B S2022 © Institute and Faculty of Actuaries

 CS1B – Actuarial Statistics – Core Principles – September 2022 - Examiners’ report

A. General comments on the aims of this subject and how it is marked

The aim of the Actuarial Statistics subject is to provide a grounding in mathematical and
statistical techniques that are of particular relevance to actuarial work.

In particular, the CS1B paper is a problem-based examination and focuses on the

assessment of computer-based data analysis and statistical modelling skills.

For the CS1B exam candidates are expected to include the R code that they have used to
obtain the answers, together with the main R output produced, such as charts or tables.

When a question requires a particular numerical answer or conclusion, this should be

explicitly and clearly stated, separately from, and in addition to the R output that may
contain the relevant numerical information.

Some of the questions in the examination paper accept alternative solutions from those
presented in this report, or different ways in which the provided answer can be
determined. In particular, there are variations of the R code presented here, which are
valid and can produce the correct output. All mathematically and computationally valid
solutions or answers received credit as appropriate.

In cases where the same error was carried forward to later parts of the answer, candidates
were given full credit for the later parts.

In questions where comments were required, valid comments that were different from
those provided in the solutions also received full credit where appropriate.

In cases where a question is based on simulations, and no seed was specified, all
numerical answers provided in this document are examples of possible results. The
numerical values presented here will be different if the simulations are repeated.

B. Comments on candidate performance in this diet of the examination.

Overall performance in CS1B was satisfactory. Well prepared candidates were able to
score highly.

Most candidates demonstrated sufficient knowledge of the key R commands required for
the application of the statistical techniques involved in this subject.

The quality of the commentary given alongside the R output was not always strong or
insufficient (e.g. in Question 2). Performance in questions with atypical style (e.g. Q4(iv))
was relatively weak, despite the tested topics being standard basic statistical concepts
from the CS1 Core Reading. This highlights the need for candidates to cover the whole
syllabus when they revise for the exam and not rely heavily on questions appearing in
recent papers.

In some cases, the layout of the provided answers was not satisfactory, with input and
output being separated and presented with large gaps, and with significant duplications.

CS1B S2022 © Institute and Faculty of Actuaries

 CS1B – Actuarial Statistics – Core Principles – September 2022 - Examiners’ report

C. Pass Mark

The Pass Mark for this exam was 55

1302 presented themselves and 539 passed.

Solutions for Subject CS1B – September 2022

Q1
(i)
set.seed(2022)

x_bar <- rep(0,5000) [1]

for (i in 1:5000) [1]
+ {x <- rpois(150,6) [2]
+ x_bar[i] <- mean(x)} [1]
mean(x_bar) [½]
[1] 5.999752 [½]
> var(x_bar) [½]
[1] 0.03958983 [½]

(ii)
From the Central Limit Theorem, the distribution of the sample means will be
approximately:
6
𝑁𝑁 �6, 150� , i. e. 𝑁𝑁(6 , 0.04) [1]

(iii)
The mean and variance from part (i) are very close to the approximate Normal
distribution in part (ii) [1]

(iv)
qqnorm(x_bar) [2]
qqline(x_bar)

CS1B S2022 © Institute and Faculty of Actuaries

 CS1B – Actuarial Statistics – Core Principles – September 2022 - Examiners’ report

[1]

(v)
The QQ plot shows that the quantiles of x_bar and the Normal distribution are very
similar [1]
Therefore, the Normal distribution is a good approximation for x_bar [1]
x_bar has a slightly lighter upper tail than the Normal distribution [1]
x_bar also has a slightly lighter lower tail than the Normal distribution [1]
The approximation will improve with larger sample sizes [1]
[Mars available 5, maximum 4]
[Total 16]

This question was answered well by most candidates.

There were some errors in specifying the variance in part (ii).

In part (iv), well prepared candidates were able to produce a QQ plot along with
reasonable comments. A common error in part (iv) was using random values rather than
the quantiles of the distribution to produce the plot.

In part (v), a relatively small proportion of candidates were able to identify and interpret
differences in the tails.

Q2
(i)
𝑋𝑋 follows a Binomial (𝑛𝑛, 𝜃𝜃) distribution, where 𝑛𝑛 is the number of clients surveyed
and 𝜃𝜃 is the probability of getting a positive approval of the product from a client [1]

CS1B S2022 © Institute and Faculty of Actuaries

 CS1B – Actuarial Statistics – Core Principles – September 2022 - Examiners’ report

(ii)
The posterior distribution of 𝜃𝜃 is Beta with parameters
𝛼𝛼 + 𝑥𝑥 = 𝛼𝛼 + 101 and 𝑛𝑛 − 𝑥𝑥 + 𝛽𝛽 = 59 + 𝛽𝛽 . [2]

(iii)
The prior and posterior are of the same family. So, we have a conjugate prior [1]

(iv)
Beta(1, 1) is the same as a Uniform(0,1) distribution: 𝛼𝛼 = 1, 𝛽𝛽 = 1 [1]

(v)(a)
n = 160
x = 101
alpha = 1
beta = 1
p = seq(0,1, by=0.01) [½]
plot(p, dbeta(p, alpha, beta), ylab="density", type ="l",
ylim = c(0,12), main = "Plot for the prior density of θ") [1]

[½]

(b)
lines(p, dbeta(p, x+alpha, n-x+beta), type ="l", col="red") [1]

legend(0.,12, c("Prior U(0,1)",

"Posterior"),lty=c(1,1),col=c("black", "red")) [½]

CS1B S2022 © Institute and Faculty of Actuaries

 CS1B – Actuarial Statistics – Core Principles – September 2022 - Examiners’ report

[½]
(vi)
alpha1 = 40
beta1 = 24

p = seq(0,1, by=0.01)
lines(p, dbeta(p, alpha1, beta1), col="blue") [1]

lines(p, dbeta(p, x+alpha1, n-x+beta1), type ="l",

col="brown") [1]

legend(0.,12, c("U(0,1)", "Beta(102,60)", "Beta(40, 24)",

"Beta(141,83)"),lty=c(1,1),col=c("black", "red", "blue",
"brown"))

[1]

(vii)
The Beta(40, 24) prior contains more information than the uniform (0, 1) prior [1]
and this is reflected in the posterior densities obtained as the posterior in (vi) is narrower
than the one in (v) [1]

CS1B S2022 © Institute and Faculty of Actuaries

 CS1B – Actuarial Statistics – Core Principles – September 2022 - Examiners’ report

(viii)(a)
Using the Uniform(0, 1) prior:

1 - pbeta(3/5, x+alpha, n-x+beta) [1½]

0.7840908 [½]

Using the Beta(40, 24) prior:

1 - pbeta(3/5, x+alpha1, n-x+beta1) [1½]

0.8199189 [½]

(b)
The Beta(40, 24) prior provides a slightly higher posterior probability than the proportion of
clients positively perceiving the new product [1]

As the difference between the two priors is considerable and the difference between the
two posteriors is small, we can conclude that most information is coming from the data rather
than the priors [1]
[Total 20]

This question was not well answered, with many candidates not attempting parts of it.

In parts requiring comments, credit was awarded for valid comments that are not shown
here.

In parts where a legend is shown on the graph, credit was given when candidates
provided a relevant description or comment in the answer instead of the legend.

In part (iv), a common error was to give (0,1) as the parameters of the distribution.

In part (viii)(a), a common error was using the prior instead of the posterior distribution.

Q3
# load data
load("mile_records.Rdata")

(i)
plot(record.date, record.time, ylab="New World Record Times
(seconds)", xlab="Days since 31.05.1913",main="Plot for
mile_records data set") [1]

CS1B S2022 © Institute and Faculty of Actuaries

 CS1B – Actuarial Statistics – Core Principles – September 2022 - Examiners’ report

[1]

(ii)
cor(record.date, record.time) [1]
-0.9885164 [1]

(iii)
model = lm(record.time ~ record.date) [1]
coefficients(model) [1]

The intercept of the regression line is 254.8 [½]

and the slope is -0.001076 [½]

(iv)
abline(model) [1]

CS1B S2022 © Institute and Faculty of Actuaries

 CS1B – Actuarial Statistics – Core Principles – September 2022 - Examiners’ report

[1]

(v)
H0: beta = 0, versus H1: beta ≠ 0 [1]
summary(model) [1]

Call:
lm(formula = record.time ~ record.date)

Residuals:
Min 1Q Median 3Q Max
-2.6103 -0.7494 -0.1780 0.6999 2.8474

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 2.548e+02 5.409e-01 471.08 <2e-16 ***
record.date -1.076e-03 3.004e-05 -35.83 <2e-16 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 1.379 on 30 degrees of freedom

Multiple R-squared: 0.9772, Adjusted R-squared: 0.9764
F-statistic: 1284 on 1 and 30 DF, p-value: < 2.2e-16

P-value is very small (i.e. p-value: < 2.2e-16) [1]

So conclude that negative slope is significant [1]

(vi)
The plot clearly shows a negative relationship between days since May 1913 and the
current one-mile world record. [1]
The correlation coefficient is very close to -1, and the slope is negative and significant,
confirming the relationship [1]

CS1B S2022 © Institute and Faculty of Actuaries

 CS1B – Actuarial Statistics – Core Principles – September 2022 - Examiners’ report

However, the response variable cannot increase with time (it is a world record).
So we cannot observe a positive relationship, meaning that the above results are to be
expected [2]

(vii)
x = record.date[32]+365*100 [1]
predict(model, data.frame(record.date=x)) [2]
# Or
coefficients(model)[1] + coefficients(model)[2]*x
181.66 seconds [1]

(viii)
twoMinuteDate = (120 -
coefficients(model)[1])/coefficients(model)[2] [1]

twoMinuteDate = twoMinuteDate - record.date[32] [1]

twoMinuteDate/365 [1]

This gives: 256.97 years. [1]

(ix)
The model is suitable for modelling the past observations, and we would also consider
this a good model for predicting records in the near future as the model is significant, 𝑅𝑅 2
is close to one and the correlation is clearly negative. We have no reason to assume that
the linear relationship will break down in the near future [1]

However, in the long run the model predicts unrealistically low times for the one mile run,
eventually predicting even negative times [1]
[Total 27]

The quality of answers given in this question was mixed.

Part (i) was well answered generally, but marks were lost for not providing appropriate
annotation on the plot.

In part (v) a number of candidates failed to state the hypotheses and explicitly state a full
conclusion.

In parts (vii), (viii), a relatively small proportion of candidates provided fully correct
answers. In parts (vi) and (ix) credit was awarded for other reasonable comments.

Q4
# load data
load("employee.RData")

(i)
Categorical variables: gender, job.type and job.location [1]

CS1B S2022 © Institute and Faculty of Actuaries

 CS1B – Actuarial Statistics – Core Principles – September 2022 - Examiners’ report

Numerical variables: salary.current, salary.start, age and experience [1]

(ii)
Scattergraphs between each pair of continous variables:

data_continuous <- data.frame(salary.current, salary.start,

age, experience) [1]
pairs(data_continuous) [1]
or
plot(data_continuous)

[1]

(iii)
The first column shows the relationship between salary.current and the other continuous
variables. We can see that:

there appears to be a positive linear relationship between the current and the starting
salary [1]
there appears to be a weak negative linear relationship between current salary and the
employee age and the employee experience [1]
age has an increasing relationship with salary at the very start but then it becomes
unclear [1]

CS1B S2022 © Institute and Faculty of Actuaries

 CS1B – Actuarial Statistics – Core Principles – September 2022 - Examiners’ report

[Marks available 3, maximum 2]

(iv)(a)
The lower quartile, median, upper quartile of salary.current are given as:

quantile(salary.current, c(0.25,0.5,0.75)) [½]

25% 50% 75%
9.1750 9.3650 9.8625 [1½]

and the mean:

mean(salary.current) [1]
# 9.4778

## Alternatively, candidates can use the function summary

summary(salary.current)
Min. 1st Qu. Median Mean 3rd Qu. Max.
8.780 9.175 9.365 9.478 9.863 10.630

(b)
We use the following notation:
p1 : proportion of male employees with current salary values below 9.86
p2 : proportion of female employees with values below 9.86

𝐻𝐻0 : p1 = p2
𝐻𝐻1 : p1 ≠ p2 [1]

Proportion calculations:

Current salary for male employees:

cmale<- salary.current[gender==0] [½]

Those less than the upper quartile:

qmale<- cmale[cmale <9.86] [1]

Number of male employees with current salary lower than the upper quartile:
x1<- length(qmale) ; x1 [1]
# 26

Number of male employees:

n1<- length(cmale); n1 [1]
# 51

Current salary for female employees:

cfemale<- salary.current[gender==1] [½]

Those less than the upper quartile:

qfemale<- cfemale[cfemale <9.86] [1]

CS1B S2022 © Institute and Faculty of Actuaries

 CS1B – Actuarial Statistics – Core Principles – September 2022 - Examiners’ report

Number of female employees with current salary lower than the upper quartile:
x2<- length(qfemale) ; x2 [1]
# 48

Number of female employees:

n2<- length(cfemale); n2 [1]
# 49

Perform test:
prop.test(c(x1,x2), c(n1 , n2 ), alternative = "two.sided" )[2]

2-sample test for equality of proportions with continuity correction

data: c(x1, x2) out of c(n1, n2)

X-squared = 26.276, df = 1, p-value = 2.959e-07
alternative hypothesis: two.sided
95 percent confidence interval:
-0.6325920 -0.3069838
sample estimates:
prop 1 prop 2
0.5098039 0.9795918

The p-value is less than 0.05, therefore reject 𝐻𝐻0 . The proportion of males with current
salary below 9.86 is significantly different from the proportion of female employees with
current salary below 9.86 [1]

(v)
The median, mean and variances of salary.current for each of the job type in job.type can
be given as follows:

Median<- rep(0,5)
Mean<- rep(0,5)
Var<- rep(0,5
for(i in 1:5){
Median[i]<- median(salary.current[job.type==i] )
Mean[i]<- mean(salary.current[job.type ==i] )
Var[i]<- var(salary.current[job.type ==i] )
} [3]

## Median
# 9.180 9.285 9.410 10.160 10.030 [1]

## Mean
# 9.168571 9.280333 9.440000 10.101333 10.119167 [1]

## Var
# 0.07471261 0.02906540 0.00320000 0.03128381 0.05988106 [1]

(vi) (a)

CS1B S2022 © Institute and Faculty of Actuaries

 CS1B – Actuarial Statistics – Core Principles – September 2022 - Examiners’ report

𝐻𝐻0 : Mean starting salaries and the current salaries are equal
𝐻𝐻1 : Mean starting salaries and the current salaries are not equal. [½]

t.test(salary.start,salary.current, paired = TRUE) [2]

#Results [1]
Paired t-test

data: salary.start and salary.current

t = -38.506, df = 99, p-value < 2.2e-16
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
-0.7190359 -0.6485641
sample estimates:
mean of the differences
-0.6838

The p-value is less than 0.05; therefore we reject 𝐻𝐻0 and conclude that the current salary
is significantly different from the starting salary [1½]

(b)
We define two vectors of current salaries for big- and small-city employees:

salary.current_bcity<- salary.current[job.location==0]
salary.current_scity <- salary.current[job.location==1] [1]

𝐻𝐻0 : Mean starting salaries for big and small city employees are equal
𝐻𝐻1 : Mean starting salaries for big city employee > small city employee [½]

t.test(salary.current_bcity, salary.current_scity,alternative=
"greater") [1½]

# Results [½]
Welch Two Sample t-test
data: salary.current_bcity and salary.current_scity
t = 5.2366, df = 90.239, p-value = 5.303e-07
alternative hypothesis: true difference in means is greater than 0
95 percent confidence interval:
0.2481034 Inf

sample estimates:
mean of x mean of y
9.572297 9.208846

We obtain a p-value<0.05, and therefore reject 𝐻𝐻0 at 5%. We conclude that the current
salary for big-city employees is greater than that for small-city employees [1½]
[Total 37]

CS1B S2022 © Institute and Faculty of Actuaries

 CS1B – Actuarial Statistics – Core Principles – September 2022 - Examiners’ report

This question was answered well in general.

The computations in parts (iv) and (v) can be performed in a number of alternative ways
in R, and credit was given as appropriate when alternative R code was presented.

Many candidates failed to provide a fully correct answer in part (iv)(b), where a number
of errors were found in the calculations.

In parts (vi)(a), (vi)(b) a number of candidates failed to state the hypotheses of the tests
and complete conclusions.

[Paper Total 100]

END OF EXAMINERS’ REPORT

CS1B S2022 © Institute and Faculty of Actuaries

Beijing
14F China World Office 1 · 1 Jianwai Avenue · Beijing · China 100004
Tel: +86 (10) 6535 0248

Edinburgh
Level 2 · Exchange Crescent · 7 Conference Square · Edinburgh · EH3 8RA
Tel: +44 (0) 131 240 1300

Hong Kong
1803 Tower One · Lippo Centre · 89 Queensway · Hong Kong
Tel: +852 2147 9418

London (registered office)

7th Floor · Holborn Gate · 326-330 High Holborn · London · WC1V 7PP
Tel: +44 (0) 20 7632 2100

Oxford
1st Floor · Belsyre Court · 57 Woodstock Road · Oxford · OX2 6HJ
Tel: +44 (0) 1865 268 200

Singapore
5 Shenton Way · UIC Building · #10-01 · Singapore 068808
Tel: +65 8778 1784

www.actuaries.org.uk
© 2021 Institute and Faculty of Actuaries